Skip to content

Conversation

@devopsdymyr
Copy link

Fix documentation typos and syntax errors in README.md

Summary

This PR fixes several typos and syntax errors found in the README.md documentation to improve clarity and correctness.

Changes Made

  • Fixed typo: maintstreammainstream (line 32)
  • Fixed typo: ShudownShutdown (line 100)
  • Fixed typo: min_repliamin_replica (line 141)
  • Fixed quote inconsistency in Gemini example: 'name': "gemini"'name': 'gemini' (line 155)
  • Added missing quotes for sglang_llm app_type: app_type: sglang_llm'sglang_llm' (line 164)
  • Fixed typos: botttleneckbottleneck and generaiongeneration (line 329)

Why?

These documentation errors can cause confusion for users following the README examples and reduce the professional quality of the documentation. Fixing these typos and syntax errors ensures:

  • Better readability and clarity for users
  • Correct code examples that won't cause syntax errors when copied
  • Improved professional appearance of the documentation

How?

All changes are straightforward typo corrections and syntax fixes:

  • Corrected spelling errors in text content
  • Fixed inconsistent quote usage in code examples to ensure proper Python dictionary syntax
  • Added missing quotes around string values in code examples

Per CONTRIBUTING.md guidelines (lines 3-13): Updated documentation to ensure clarity and correctness.

Test Plan

  • Reviewed all changes via git diff to ensure accuracy
  • Verified markdown syntax is valid
  • Confirmed code examples maintain proper Python dictionary syntax
  • Checked that all line references are correct

Type of Change

  • Documentation fix
  • Bug fix
  • New feature
  • Breaking change

Checklist

  • Forked the repo and created branch from main
  • Updated documentation (README.md)
  • Code follows project style guidelines
  • Added tests (N/A - documentation only)
  • Updated API documentation (N/A - no API changes)

Contributor License Agreement

I understand that a CLA may be required for this contribution. I will complete it if requested by the maintainers.

- Fix typo: maintstream -> mainstream
- Fix typo: Shudown -> Shutdown
- Fix typo: min_replia -> min_replica
- Fix quote inconsistency in Gemini example (name field)
- Add missing quotes for sglang_llm app_type
- Fix typos: botttleneck -> bottleneck, generaion -> generation

Per CONTRIBUTING.md (3-13): Updated documentation to ensure clarity and correctness.
@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Dec 3, 2025
…acebookresearch#111)

- Add get_gpu_requirements_per_replica() to calculate GPU needs per app
- Add sort_apps_by_gpu_requirements() to sort apps by GPU requirements (largest first)
- Integrate sorting into deploy() method when deploying multiple apps
- Add comprehensive unit tests for fragmentation fix

This prevents resource fragmentation by deploying models requiring more GPUs
first, leaving smaller contiguous blocks for models requiring fewer GPUs.

Fixes facebookresearch#111
logger.info(
"Sorting applications by GPU requirements (largest first) to minimize fragmentation"
)
applications = sort_apps_by_gpu_requirements(applications)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

not sure if sorting makes a difference?

Existing deployment could partially occupy a node, which prevent later deployment to get resources without removing old one. Not sure if Ray also does a poor job of compaction for multiple models in the same deployment.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants